A Study Of Welding Quality Control System Based On Multi-sensor Information Fusion

Welding quality directly affects the performance of weldments, welding defects may lead to the weakening of the strength of welded joints, so as to cause its breaking during use. At present, welding robots used in our country are most of teaching and playback robots or some semi-automatic welding robots; but due to the influence of environment and the assembly condition, most welding robots used in our country can not meet the requirement of high quality of welding. Although there are lots of tests has been done by scholars in our country on the study of control of welding quality in welding process, these studies and tests are mostly related to the accuracy of seam tracking, researches about welding puddle, which has the same impact on the quality of welding, especially on the monitoring of welding penetration in real-time are slightly less.According to those shortages, this paper first propose a method by using arc sensor and ultrasonic sensor fusion information to get the prediction model of welding penetration. To develop the control of welding quality online, a welding quality monitoring and controlling system based on a set of multi-sensor information fusion is presented in this paper.On the basis of previous researches, the model of welding arc and welding power is improved in this paper: the changing height of the torch is introduced in welding arc length module of welding arc model, the output voltage module of welding power model is replaced by control voltage waveform, the simulation result has shown that these improvements make the arc sensing system model closer to the actual system. And then a method of extracting feature information of welding seam deviation from welding arc current information by using the arc sensing system model is presented in this paper. A study on welding penetration measurement using ultrasonic was also conduted in this paper and a simplified ultrasonic weld penetration measurement model was built to study the relationship of ultrasonic and welding penetration, then the welding penetration prediction model was built by using BP neural network. To develop the model, the BP neural network system fuses the welding seam deviation information extracted from arc sensor observating information, welding current information, welding speed information, groove type information and ultrasound sensor observating information to built the welding penetration predition model. Finaly, a parameters self-tuning fuzzy PID controllers is designed based on seam tracking and weld penetration control to monitor and control the process of welding. The result shows that the designed welding quality monitoring and controlling system based on multi-sensor information fusion can effectively improve the quality of welding.